OpenAI, DeepSeek, and Anthropic Release New AI Models in Rapid Succession

The pace of AI model releases is accelerating. Just days after Anthropic launched Claude Opus 4.7, OpenAI has unveiled GPT-5.5, and China’s DeepSeek dropped a preview of its V4 series. Each company is angling for a bigger slice of the enterprise and consumer markets, pouring billions into new capabilities.

DeepSeek’s V4 Flash and V4 Pro models focus on reasoning and agentic tasks, using a Hybrid Attention Architecture that retains context across lengthy conversations. The company also claims architectural improvements allow the models to run on cheaper hardware, potentially lowering deployment costs for businesses. DeepSeek disrupted the industry last year with low-cost models that matched top-tier performance, but the company now faces political headwinds: the White House this week accused Chinese AI firms of widespread technology theft.

OpenAI, which has accused DeepSeek of copying its models, rolled out GPT-5.5 to paying ChatGPT and Codex subscribers. The update emphasizes coding, computer use, and research. President Greg Brockman told reporters the model is more intuitive and requires less human guidance. “It can look at an unclear problem and figure out just what needs to happen next,” he said.

Anthropic’s Opus 4.7, meanwhile, is positioned as a safer, more literal alternative to its powerful but unreleased Mythos model, which the company considers too risky for cybersecurity. Opus produces cleaner outputs and follows prompts more directly, making it useful for developers who need straightforward results.

Source: CNET